Image segmentation for identifying and tracking objects in images and videos is required for a variety of applications such as medical imaging, remote sensing, security surveillance and military target detection. One method of tracking image objects is through active contours whose evolution on an image plane leads to image segmentation and object tracking.
Typically, an energy functional is defined on the contour, based on either region or gradient properties of the model. Further, the total energy functional is estimated as a weighted combination of internal and external energy terms. In some cases, a combination of both properties is used. The minima of the total energy functional is obtained through solution of Euler Lagrange equations for the energy functional. Typically, for the energy functionals defined on the contour, these are generally partial differential equations and are solved numerically till convergence. Furthermore, for gradient based active contour models, clutter surrounding the image object may lead to multiple local minima thereby resulting in incorrect segmentation.